Algorithm Development in Materials Science and Engineering: Large Scale Computational Simulations and Microscale Algorithms for Study Structure-Processing Relations
Sponsored by: TMS Materials Processing and Manufacturing Division, TMS: Computational Materials Science and Engineering Committee, TMS: Integrated Computational Materials Engineering Committee, TMS: Phase Transformations Committee, TMS: Solidification Committee
Program Organizers: Mohsen Asle Zaeem, Colorado School of Mines; Mikhail Mendelev, NASA ARC; Bryan Wong, University of California, Riverside; Ebrahim Asadi, University of Memphis; Garritt Tucker, Colorado School of Mines; Charudatta Phatak, Argonne National Laboratory; Bryce Meredig, Travertine Labs LLC

Tuesday 8:30 AM
March 16, 2021
Room: RM 36
Location: TMS2021 Virtual

Session Chair: Cheikh Cisse, Honeywell; Mohsen Asle Zaeem, Colorado School of Mines


8:30 AM  
Exascale-motivated Algorithm Development for Nano and Mesoscale Materials Methods: Sam Reeve1; Matthew Rolchigo1; Jim Belak1; 1Lawrence Livermore National Laboratory
    With exascale computing on the horizon, new hardware and software are enabling simulations across all materials simulation methods. The ability to fully leverage these machines requires significant redesign of algorithms (mainly for GPUs), but has also increasingly required performance portability such that codes work across hardware. We describe the materials-relevant, algorithm-driven efforts of two parts of the Exascale Computing Project: CoPA for particle applications and ExaAM for additive manufacturing. These efforts span scales and physical processes, from the nanoscale with both quantum mechanical (QM) and classical molecular dynamics (MD), to mesoscale methods including particle-in-cell (PIC) for structural mechanics and cellular automata (CA) for microstructure evolution. Across the methods, emphasis has been placed on development of software libraries (for separation of concerns from the materials scientist), leveraging hardware vendor libraries and the Kokkos programming model. Other major themes include methods for adoption of portability libraries and to reduce communication.

8:50 AM  
Preparing for Exascale Phase-field Simulations: Scalable, Performance-portable Precipitation Simulations: Stephen DeWitt1; Philip Fackler1; Younggil Song1; Bala Radhakrishnan1; John Turner1; 1Oak Ridge National Laboratory
    Exascale computing can enable 3D, full-physics simulations of phenomena at larger ranges of length and time scales than is currently possible. However, developing performant codes for leadership machines with a variety of heterogeneous-node architectures poses a significant challenge. We discuss an approach taken to tackle this challenge in MEUMAPPS, an FFT-based phase-field code. Since the best performing FFT library might differ across machines, the code is designed for easy integration with different scalable FFT libraries. The code leverages Kokkos for portability across node-level parallelization methods (e.g. CUDA, HIP, OpenMP). The results of GPU-speedup, scaling, and profiling studies for simulated precipitation in Ni-base superalloys are reported. This manuscript has been authored by UT-Battelle, LLC under Contract No. DE-AC05-00OR22725 with the U.S. Department of Energy. Research supported by the Exascale Computing Project (17-SC-20-SC), a joint project of the U.S. Department of Energy’s Office of Science and National Nuclear Security Administration.

9:10 AM  
Tusas: A Modern Computational Approach for Microstructure Evolution Toward Exascale: Supriyo Ghosh1; Christopher Newman1; Marianne Francois1; 1Los Alamos National Laboratory
    Predictive microstructure evolution in materials during additive manufacturing is a critical challenge problem for the exascale computing project. We address this challenge using a multi-GPU/CPU-based hybrid implementation of quantitative phase-field methods for alloy solidification. We use our in-house Tusas code for solidification simulations. Tusas is a general, flexible, open-source tool developed in C++ with advanced numerical algorithms using finite-element discretization for unstructured meshes and is optimized to run efficiently on massively parallel heterogeneous supercomputers. Our computational approach is analyzed using appropriate validation and verification models, algorithmic performance, parallel scalability, and large-scale long-time solidification simulations. Finally, we characterize the evolution of microstructure characteristics, such as cellular spacing and solute segregation, from the solidification simulations relevant to exascale applications. (LA-UR-20-24954)

9:30 AM  
Bayesian Data Assimilation for Phase-field Simulation of Solid-state Sintering: Akimitsu Ishii1; Akinori Yamanaka1; Yuki Okada1; Akiyasu Yamamoto1; 1Tokyo University of Agriculture and Technology
    Phase-field (PF) method is a powerful numerical simulation methodology for analyzing microstructural evolutions during a solid-state sintering. However, many physical values and material parameters of sintered materials, such as polycrystalline bulk superconducting materials, are unknown and immeasurable. On the other hand, recently, the three-dimensional (3D) microstructural evolution during the sintering can be observed using advanced experimental techniques (e.g. X-ray computed tomography). Data assimilation (DA) based on the Bayesian inference enables us to combine the experimental data with the numerical simulation and to identify unknown physical values and material parameters. In this study, we have applied an ensemble-based four-dimensional variational (En4DVar) DA method to a 3D phase-field simulation of solid-state sintering. Through numerical experiments, we show that En4DVar can simultaneously estimate multiple material parameters including a grain boundary mobility only from the morphological data of sintered material.

9:50 AM  
Phase Field Dislocation Dynamics (PFDD) Modeling of Non-Schmid Effects in BCC Metals: Hyojung Kim1; Nithin Mathew1; Darby J. Luscher1; Abigail Hunter1; 1Los Alamos National Laboratory
    Screw dislocations determine the plastic deformation of body-centered cubic (BCC) metals. The critical resolved shear stress (CRSS) of BCC metals deviates from the Schmid law, indicating that the Peierls barrier is dependent on the stress state. We account for non-Schmid behavior in Phase field dislocation dynamics (PFDD), which describes the total energy of dislocation configuration with elastic strain energy, core energy, and external energy, in two ways. One way is to incorporate non-glide stress components in the external energy. However, this method still produces discrepancies when the maximum resolved shear stress plane is the {110} glide plane, indicating that our understanding of how non-Schmid behavior affects overall material response is not complete. Alternatively, we suggest accounting for the non-Schmid behavior in the core energy by variation of molecular statics (MS)-computed stacking fault energy. The CRSS predicted using PFDD and comparison to MS predictions will be discussed.

10:10 AM  
A Quantitative Phase-field Model for Study of Shape Memory Behavior and Elastocaloric Effect in CuAlBe: Cheikh Cissé1; Mohsen Asle Zaeem1; 1Colorado School of Mines
    We are presenting a quantitative non-isothermal elastoplastic phase-field model for shape memory alloys. It is fully thermo-mechanically coupled considering the effects of temperature-dependent properties, latent heat, grain boundaries, and asymmetric transformation and plasticity. The asymmetric plasticity formulation is added to this model in a more realistic way than the TDGL-like diffusive plasticity which is also computationally more efficient than the crystal plasticity. The proposed model is the first of its kind to capture, beyond shape memory effect and pseudoplasticity, the stress-assisted two-way memory effect, the formation of internal loop during thermal cycling, and thermomechanical training. Comparison between tension and compression demonstrates the capability of this model to account for nonsymmetrical transformation and plastic responses of shape memory alloys. The elastocaloric effect related to the heat release and absorption during phase transformation is simulated and the microstructure correlated with the temperature distribution. Ultimately, the elastocaloric performance in functional fatigue is assessed.